Curiosity.
With context.

01 / Orientation

A map, with room for nuance.

Roast moves from light at the bottom to dark at the top. Flavor moves from cocoa and nut-led profiles toward fruit and florals. These positions help you explore; they don’t rank quality.

Roast descriptions come from the cited source. Flavor positions are coarse editorial interpretations of attributed notes, recorded in five-point steps. Spice, herbal, earthy, and other characteristics deserve more room than one axis can give them, so each review explains its placement and keeps the tasting notes.

The vertical axis maps source labels to five bands: light (0), medium-light (25), medium (50), medium-dark (75), and dark (100). These are navigation positions, not Agtron readings or measurements of roast intensity. To reveal overlapping profiles, the display uses softly tapered, stable offsets tied to each coffee’s ID. Documented roast positions stay within their broad bands; unknown roasts receive a wider spread around the medium-light assumption. These offsets do not add measured precision to the underlying source values.

Hover to preview, click or tap to pin, then open a review or save a coffee. Zoom and pan to explore nearby profiles. Coffees that share a position are individually accessible in the results and through keyboard navigation. Hollow dots have an estimated roast position: when a source does not specify roast, we use a low-confidence medium-light starting assumption. Switch estimates off to see only profiles with documented roast bands. Flavor positions still require attributed notes. Roast filters use documented labels, not estimates. Contours show concentrations only when there are enough records; they never indicate quality.

02 / Assessment

There’s a record behind the number.

Open a score to see its source, date, method, and individual categories. A reviewer rating stays on that reviewer’s scale. A roaster’s reported cupping stays attributed to that roaster. “Our cupping” is reserved for documented firsthand assessments.

Scores from different methods are not averaged into a universal rating. An acidity quality score, for example, is different from how intensely acidic a cup tastes.

We preserve a source’s published numbers. Missing categories stay missing, and a score for an old crop or a different roast never automatically transfers to a new coffee.

For more on descriptive and affective assessment, visit the Specialty Coffee Association’s Coffee Value Assessment.

Editorial estimates

A profile worth exploring.

Every researched coffee also receives an editorial-interest rating out of five. It is a reproducible estimate of the documented profile’s discovery interest, not a prediction of cup quality or a replacement for cupping.

Each of five criteria earns one point: at least three distinct attributed descriptors, at least two recognized flavor families, documented processing, a named region or producer, and a named variety. The review shows every criterion, including those not met. Missing documentation lowers the rating; it does not establish a worse coffee. Price, roast darkness, species and brand never earn points.

The sourcing preview and website apply the same versioned rubric to the flat-file record. There are no random scores, invented tastings, or extra decimal precision. These estimates stay separate from sourced quality assessments and are excluded from structured review ratings and the sourced-score filter.

03 / Editorial context

Clear about what we know.

The library combines original tasting records, supplier assessments, and attributed published reviews. Each entry identifies its basis. Curated summaries are not presented as firsthand tasting.

Agents help organize source material and check the content. Sensory judgments and cupping scores originate with people who actually tasted the coffee. Unknown dates, crop years, and lot identifiers remain unknown.

Some reviews describe historical offerings. A sourced score applies only to the reviewed sample; buying links identify their last check date.

Researched reviews can have no published score. They still offer attributed notes and original editorial context. A missing rating does not imply low quality, and it is never replaced with an AI-generated tasting score.

A note on imagery

Images with context.

Verified photographs identify individual offerings. Where no photograph is cleared for use, an original illustrated profile cover identifies the coffee by name, roaster, and origin. Its abstract curves are decorative, not measurements, packaging or photographs of the sample. Packaging, origin photographs and the reviewed sample have different meanings; we make that relationship explicit. Current packaging does not establish that an older assessment applies to today’s lot.

04 / Remembering a discovery

A little collection of your own.

Save a coffee from the map or its review. Your shortlist stays in this browser without an account. Clearing browser storage clears the list, and it does not sync to other devices.